A phosphorus-31 solid-state nuclear magnetic resonance study of

Jun 26, 1990 - Uniuersity of Wisconsin—Madison, Madison,Wisconsin 53706. Philip E. Pfeffer ... Alabama A&M University, Normal, Alabama 35762...
0 downloads 0 Views 1MB Size
1702

Langmuir 1991, 7, 1702-1712

A 31PSolid-state Nuclear Magnetic Resonance Study of Phosphate Adsorption at the Boehmite/Aqueous Solution Interface William F. Blea" University of Wisconsin-Madison, Madison, Wisconsin 53706

Philip E. Pfeffer USDA-ARS Eastern Regional Research Center, Philadelphia, Pennsylvania 19118

Sabine Goldberg USDA-ARS US.Salinity Laboratory, Riverside, California 92502

Robert W. Taylor Alabama A&M University, Normal, Alabama 35762

Robert Dudley USDA-ARS Eastern Regional Research Center, Philadelphia, Pennsylvania 19118 Received June 26,1990. In Final Form: January 31, 1991 Results from 31P solid-state NMR experimentsprovide direct evidencefor the hydrolysis of phosphate molecules adsorbedon the surfaceof boehmite (yAl00H). Much of the phosphate surface-excessmolecules are in chemical environmentsthat are highly sensitive to solution pH, and the phosphate molecules are held relatively immobile. This chemical environment is believed to be inner-sphere surface complexes. The resultsof "interrupted-decoupling"3lP(lHJcross-polarization,magic-angle-spinningNMR experiments are used to discriminate protonated surface phosphate species from those that are fully deprotonated. The pH at which surfacephosphates becomedeprotonatedlies somewhere between 9 and 11. Thistransition pH is an NMR observable which can be predicted by surface-complexationmodels.

Introduction The environmental consequence of phosphate pollution occurs mainly in surface waters. Here high levels of bioavailable phosphate, often in concert with high concentrations of dissolved nitrate, stimulate blooms of aquatic vegetation. These blooms deplete dissolved oxygen,killing fish and other aquatic animals. Phosphate bioavailability appears to be controlled through some combination of phosphate-mineral solubility, degradation of organic phosphate compounds, and adsorption at oxide mineral/ aqueous solution interfaces. This study examines phosphate adsorption at the aluminum oxide/aqueous solution interface through modeling and 3lP solid-state nuclear magnetic resonance (NMR). Among other things, the surface-complexation model predicts pH-dependent hydrolysis of adsorbed phosphate molecules. The solid-state nuclear magnetic resonance experiments we used monitor hydrolysis of adsorbed phosphate and, if it occurs, precipitation. A t issue is whether the picture that emerges from NMR experiments agrees with the representation of the surface-complexation model. Both precipitation and adsorption can occur when phosphate is added to an aluminum oxide suspension. It is probably impossibleto determine where adsorption ends and precipitation begins from measurement of solution ion concentrations alone.'-' This transition point defines ~

(1) Mattaon, S.;A l v d e r , E.; Koutler-Anderseon,E.;Vahtras,K. Ann. R. A&. Coll. Swed. 1958, 20,19. (2) Kittrick, J. A.; Jackson, M. L. J. Soil Sci. 1957, 7, 81.

0743-7463/91/2407-1702$02.50/0

the adsorbing capacity of the oxide since further increases in the phosphate concentrationmust includeprecipitation. We also examine the design of an adsorption experiment that minimizes the likelihood of heterogeneous precipitation while maximizing the amount of phosphate adsorbed. 1. Adsorption Mechanisms. Indirect experimental evidence indicates the principal mechanism of phosphate adsorption on aluminumoxide is ligand exchangeresulting in inner-sphere surface complexes. The constant-capacitance model is the simplest surface-complexationmodel that assumes a ligand-exchange mechanism for specific inner-sphere adsorption of ions at the oxide/aqueous solution interface. This model provides an excellent quantitative description of phosphate adsorption on aluminum and iron oxides.8 It is expected that more complex surface-complexation models of inner-sphere adsorption, such as the triple-layer model, would also be appropriate. An advantage of these more complex models is their abilityto represent the ionic-strengtheffectson adsorption. However,inner-sphereadsorption is virtuallyindependent of ionic strength; such is the case for phosphate adsorption on aluminum oxide in NaCl and KCl solutions ranging (3) Heu, P.H. Soil Sci. Soc. Am. h o c . 1964,28,474. (4) Hsu, P. H. Soil Sci. 1965, 99, 398. (5) Veith, J. A * Sposito, G. Soil Sci. SOC.Am. J. 1977,41,697. (6) van R i e d i j k , W. H.; Lyklema, J. Colloids Surf. 1980, I , 33. (7) Elprince, A. M.; Sposito, G. Soil Sei. SOC.Am. J. 1981, 46,277. (8)Goldberg, 5.;Sposito,G. Commun. Soil Sci. Plant A d . 1986,16, 801.

0 1991 American Chemical Society

Langmuir, Vol. 7, No. 8, 1991 1703

Phosphate Adsorption on Boehmite from 0.002 to 0.2 mol L4.9 More complex models were not used in our study because the large increase in the number of adjustable parametera was unnecessary and therefore unacceptable. Direct spectroscopicevidence of inner-sphereadsorption has proven elusive.8 The first objective of our study was to investigate the properties of phosphate adsorbed at the boehmitesurface by using 3lP nuclear magnetic resonance spectroscopy. We designed our experiments to test the following hypothesis: phosphate adsorbed on the surface of boehmite hydrolyzes with varying pH in a manner represented by the constant-capacitance model.'*l2 To test this hypothesis, we combined two approaches: representation of phosphate surface speciation using the constant-capacitance model and execution of NMR experiments designed to discriminate protonated from nonprotonated phosphate species on the oxide surface. Whether adsorbedphosphate forms inner- or outer-sphere surface complexes can be determined from the characteristics of the NMRsignalsin certain routine experiments. Our major tool is high-resolution solid-state NMR spectrosc~py.~*~~ Direct measurementsof lHdlP nuclear magneticdipole interactions employ variations of the crosspolarization (CP)experimentof Pines et al." For instance, Rothwell et a1.18 and Tropp et al.lg made qualitative comparisons of CP magic-angle-spinning (CP-MAS) and single-pulse-excitation,magic-angle-spinning (MAS) spectra. The 31Presonances more strongly "coupled" to lH nuclei show a signal enhancement in CP-MAS spectra relative to MAS spectra. Aue and co-workersmS1expanded this approach by including "dipole suppression", also known as "interrupted-decoupling"?2CP-MAS experiments. The 31Presonances more strongly "coupled" to lH show a suppression in the "interrupted-decoupling" CPMAS spectra relative to MAS spectra. 2. Boehmite Adsorption Capacity. Measuring the phosphate adsorption maximum on an oxide mineral surface was another objective of this study. Phosphate is one of many molecules capable of both adsorbing at a mineral interface and forming a mineral with components of the adsorbent. Unfortunately, the transition from adsorption to precipitation is ill-defined with the consequence that the adsorption capacity itself is ambiguous. We believe that a means for resolving this dilemma is embodied in the observation by Hsu and Rennie23as they (9) Helyar, K. R.; Munne, D. N.; Burau, R. G. J. Soil Sci. 1976,27,307. (10) Hohl, H.; Sigg, L.; Stumm, W. In Particulates in Water; Kavanaugh, M. C., Leckie, J. O., Eds.; Advances in Chemistry Series 189; American Chemical Society Washington, DC,1980; p 1. (11) Stumm, W.; Kummert, R.; Sigg, L.Croat. Chem. Acta, 1980,63, 291. (12) Sigg, L.; Stu", W. Colloids Surf. 1981, 2, 101. (13) Andrew, E. R. Int. Rev. Phys. Chem. 1981,1,195. (14) F'yfe, C. A.; Bemi, L.; Clark, H. C.; Curtin, D.; Davis, J.; Drexler,

D.; Dudley, R. L.; Gobbi, G. C.; Hartmen, J. 5.;Klinoweki, J.; Hayes, P.; Lenkinski, R. E.;Lock, C. J. L.; Paul, I. C.; Rudin, A.;Tchir, W.;Thomae, J. M.;Wasylishen, F. R. S.; Wasylishen, R. E. Philos. T r o ~R.. Soc., A 1982,306,591. (15) Fyfe, C. A.; Bemi, L.; Clark,H. C.; Davis, J. A.; Gobbi, G.C.;

Hartmen, J. S.; Hayes, P. J.; Wasylishen, R. E. In Inorganic Chemistry: Toward the 21st Century; Chisholm, M. H., Ed.; American Chemical Society Wtwhington, DC,1983; p 406. (16) Kirkpatrick, R. J.; Smith,K. A.; Schramm, S.; Turner, G.;Yaug, W.-H. Annu. Rev. Earth Planet. Sci. 1986,19,29. (17) P i s , A.; Gibby, M. G.;Waugh,J. S. J. Chem.Phys. 1978,SS,669. (18) Rothwell, W. P.; Waugh, J. S.; Ysinowski, J. P. J. Am. Chem.

1980,102,2637. (19) Tropp, J.; Blumenthal, N. C.; Waugh, J. S. J. Am. Chem. SOC. 1983.105.22. (20)Aue, W. P.; Roufoeee, A. H.; Glimcher, M. J.; Griffin, R. G. Biochemutry 1984,23,6110. (21) Roufo", A.H.;Aue, W. P.;Roberta, J. E.; Glimche,M.J.; Griffii, R.G. Biochemistry 1984,23,6116. (22) Opella, S. J.; Frey, M. H. J. Am. Chem. SOC.1979,101,5864.

wrote of phosphate adsorption on aluminum oxides: "the adsorption capacity at various pH values can be compared only when all solutions are saturated with aluminum phosphate". Referring to this as the Hsu-Rennie condition, we believe it is a prescription for measuring maximum adsorbate surface excess that is tied to mineral solubilities and solution speciation. It stands in contrast to the Langmuir method for determining the phosphate adeorption maximum." Though the Langmuir equation has a mathematical form suitable for modeling adsorption isotherms (viz., the function approaches a limiting value asymptotically), the validity of physical interpretations assigned to the parameters has been strongly questi~ned.~ Application of the Hsu-Rennie condition is most practically achieved by using the "concentration-null" method for measuring adsorbate surface excess. In the "concentration-null"method, one determines the amount of the adsorbed component which has to be addedto return the solution to the original composition.26 In our case this original solution composition is determined by mineral solubilities. Because we are adsorbing phosphate on the surface of boehmite (yAlOOH), the original solution compositions in our pH-dependent phosphate adsorption experiment are in solubility equilibrium with boehmite. The phosphate concentrations in the original solutions depend on the solubilityof the aluminumphosphate likely to form in these suspensions. The Hsu-Rennie condition is satisfied when at slight aluminum phosphate undersaturation, repeated additions of phosphate yield no further decrease in the solution phosphate concentration.

Materials We prepared synthetic boehmite by using the following modification of the procedure reported by S a n d et aler This was done by weighing 4.25 g of Al metal powder into a 1-L Erlenmeyer flask,adding 200 mL of 0.5 m o l L-l HgCla solution, and stirring for 10 min. This allowed a surface amalgam to form which eliminates oxide impurities on the surface of the metal that could nucleate phases other than boehmite. The metal powder was weshed thoroughly with distilled,deionized water, then 12.5 mL of glacial acetic acid and 50 mL of distilled, deionized water were added. The mixture was stirred while heated at 75 "C until Hz gas evolved and allowed to stand at room temperature overnight. Boilingat constant volume (maintaining volume withdistilled, deionized water) for about 8 h drove off excess acetic acid and allowed crystal nuclei to form. The reaction preparation was then heated in a preseure veeeel at 170 "C for 8 h to form the boehmite. There will still be some acetate adsorbed to the boehmite and we removed this by freeze-drying the colloidal suspension and then resuspending the boehmite in 250 mL of 30% HzOz. This suspension was boiled at constant volume for 6 h (maintaining volumewith dietilled, deionized water) to remove all traces of acetate. Finally, the colloidal boehmite suspension was freeze-dried for storage. The aluminum metal was a 99.99% purity, 100 mesh powder from Aldrich Chemical Co., Inc.; all other chemicals were reagent grade. We prepared several boehmite batches, combining them at the end of the synthesis. The specific surface area was 187.2 m2/g, determined by Porous Products, Ithaca, NY,using the standard multipoint BET procedure (NO was the adsorbing gas). Scanning electrode microscopy revealed that the boehmite crystals were fibers 0.4-0.6 pm in length and about 0.06 pm in diameter. Results of powder X-ray diffraction analysis using Cu Ka radiation (A = 1.5405 A) are shown in Table I (cf. PDF 7-324,

SOC.

----.---.

(23) Heu, P. H.; Rennie, D. A. Can. J. Soil Sci. 1962,42, 197. (24) Oleen, S. R.; Watanabe, F.S. Soil Sci. SOC.Am. hoc. 1967,21, 144.

(25) Nunn, C. C.; Evertt, D. H. J. Chem. Soc., Faraday %M. 1 1985, 79, 2953. (26) Santos, P. S.; Santos, H. 5.;Nevee, R. F. 1985,Sth International Clay Conference, Denver, CO, 1985.

Bleam et al.

1704 Langmuir, Vol. 7, No.8, 1991 Table I. Powder X-ray Diffraction Data for Synthetic Boehmite. Swcings [dl in Angstroms synthetic boehmite boehmite0 d

1

d

1

6.17 3.18 2.36 1.99 1.86 1.85 1.78 1.66 1.53 1.45

100

6.11 3.164 2.346 1.980 1.860 1.850 1.770 1.662 1.527 1.453 1.434

100 65

1.412

2 2

1.39

71 55 4 42 42 4 24 13 26

13

1.31 39 a Powder Diffraction File No. 21-1307.

1.396 1.383 1.369 1.312

55 6

30 25 6 14 6 16 10 6 2

16

PDF 20-11, PDF 5-355, and PDF 10-173). No effort was made to determine whether poorly crystalline phases were present.

Experimental Methods 1. Preparation of Suspensions and Measurement of Proton Surface Charge. We used stock suspensions containing 6250 m2 of boehmite L-1 to prepare 1OOO ma L-l suspensions for proton surface charge measurement and 250 m2L-l suspensions for both microelectrophoresis and the phosphate-adsorption experiments. All suspensions contained 0.001 mol L-l KCl as a background electrolyte, the pH being adjusted by tenth molar standardized solutions of either KOH or HCl. We used KHPO, (KDP) as our source of phosphate and AlCls as the source of aluminum. All suspensionswere prepared with Copfree distilled, deionized water and stored in polyethylene bottles under N2 gas. 2. Analytical Techniques. To measure proton surface charge, we prepared 25 mL of a lo00 mpL-' suspension, adjusted the suspension to pH = 3, and titrated the suspension, in 0.25 pHunitsteps,uptopH = 11. Weusedatenthmolarstandardized KOH solution for the titration. This was repeated by adjusting the pH of the suspension to 11with KOH followedby incremental titration to pH 3 using a standardized tenth molar HCl solution. At each increment, the pH was maintained until the rate of change is less than 0.01 pH unit per minute. The proton surface charge experiment employed a Radiometer System ETS822 autotitratator. We used a Corning Model-150 pH/Ion meter for pH measurements in all other experiments. Microelectrophoresis measurements were taken with a Pen Kem Model-501 Lazer Zee Meter. We analyzed total phosphate and total aluminum spectrophotometrically, using the Murphy-Riley methodn for phosphate analysis and the method of Bloom et al." for aluminum. Solution proton-decoupled slP NMR spectra were obtained on a JEOL JNM-GX-400 spectrometer operating at 9.4 T (viz., 161.9 MHz resonance frequency). Proton-decoupled SIP spectra were obtained on a Bruker MSL-300 NMR spectrometer operating at 7.0T (viz., alp resonance frequency of 121.49 MHz). The 90° pulse width for the JNM-GX-400 spectrometer was 22 ps for The 90° pulse width for the MSL-300 spectrometer was 5 ps for both lH and slP and spinning frequencies were approximately 3 kHz. 3. Adsorption Experiments. All suspensions in our adsorption experiments contained 250 ma of boehmite per liter. We used values for the equilibrium ion activity products for the boehmite solubility and the hydrolysis of aluminum and orthophosphate ions from Lindsay.% Lindsay does not list activity products for the formation of aluminum-phosphate solution complexes, (27) Murphy, J.; Riley, J. P. Anal. Chim. Acta 1962, 27, 31. (28) Bloom, P. R.; Weaver, R. M.; McBride, M. B. Soil Sci. SOC.Am. J. 1978, 42, 713. (29) Lindeay, W. L. Chemical Equilibria in Soils; Wiley: New York, 1979.

yet there is abundant evidence for suchspecies.We estimated their formation by using constants listed by Novozamsky and Beek.38 Finally, we chose to use amorphous aluminum phosphate as the phase determining the upper limit on phosphate solubility. Our calculations employed the activity product for amorphous aluminum phosphate reported by Veithand Sposito.m Minimum phosphate solubility, correspondingtoa total solution phosphate concentration of about 0.4 mmol L-l, occurs at pH =4.5 in such a system. We chose to use a solution concentration of 0.1 mmol L-l throughout the pH range of 3-1 1because preliminaryexperiments showed that the adsorption capacity of our boehmite suspensions above pH = 5 and below pH = 4 was too low to reliably measure significant concentration changes in solutions containing 1mmol L-l phosphate and above. The solubility and amorphous aluminum phosphate rapidly exceeds that concentration above and below the pH of minimum solubility. To perform the phosphate adsorption experiment we began by preparing replicate 250 m2 L-l boehmite suspensions. We adjusted the pH of each suspension to the desired value and immediately added sufficient AlCb solution to bring the SUBpension into solubility equilibrium, continuing to adjust the pH until stable. We used a Radiometer System ETS822 autotitratator to maintain pH-stat. After 24 h, we readjusted the pH when necessary and added sufficient KDP solution to yield an initial total phosphate concentration of 0.1 mmol L-l and maintained the pH a t the chosen value. After 24 h, we measured the pH of one set of replicate suspensions (we performed at least three replicates for each treatment), centrifuged each suspension at 15 OOO rpm on a Sorvall centrifuge for 20 min, filtered the supernatant through a membrane filter (Nuclepore Corp., polycarbonate, 0.2 pm pore diameter), and analyzed for total soluble phosphate. If the soluble phosphate concentration after adsorption was lower than 0.1 mmol L-l, we added KDP to return the phosphate concentration to 0.1 mmol L-l in the remaining sets of replicates, adjusting the solution pH if necessary. Sampling of a second set of replicates followed this second addition by 24 h (adsorption kinetic studies showed that the phosphate concentration in solution did not change after a period of 3 h following the addition of KDP to the Suspension). If the soluble phosphate concentration was, once again, lower than the initial 0.1 mmol L-l concentration, we added KDP to return the phosphate concentration to 0.1 mmol L-l in the remaining set of replicates and took samples as just described 24 h later. The 'concentration null" method requires that the initial and the final phosphate concentration be equal. The boehmite suspensions were undersaturated with respect to amorphous aluminum phosphate. Near the pH of maximum adsorption, we had to make three additionsto KDP before the adsorbingcapacity of the boehmite was saturated. Two additions were sufficient when the adsorption maximum fell below =l pmol m-2. For slP solid-state NMR analysis, we prepared suspensions and performed adsorption experiments as described above. We reserved the solids after the tinal step and freeze-dried them. We prepared suspensions for microelectrophoresis in the same way that we prepared them for adsorption experiments. The 'pristine" samples, of course, did not receive any KDP. Because light scattering was too great in dispersed suspensions to measure mobilities, we centrifuged for 20 min at 15 OOO rpm and resuspended a small fraction of the boehmite in the supernatant solution before measuring the zeta potential. (30) Jameson,

R.F.; Salmon, J. E. d. Chem. SOC.1964, 4013.

(31) Salmon, J. E.; Wall, J. G.L. J. Chem. SOC.1958,1128.

(32) Akitt, J. W.; Greenwood,N. N.; Lester, G. D. J. Chem. SOC.A 1971, 2450. (33) Karlik, S. J.; Elgavish, G.A.; Pillai, R. P.; Eichhom, G. L. 1982, J. Magn. Reeon. 1982,49, 164. (34) Karlik, S. J.; Elgavieh, G.A.; Pillai, R.P.; Eichhorn, G.L. J.Am. Chem. SOC.1988,105,602. (35) Wilson, M. A.; Collins, P. J.; Akitt, J. W. Anal. Chem. 1989,61, 1253. (36)Novozamsky,I.;J.BeekInSoilChemistryPartA: &leicElemente; Bolt, G. H., Bruggenwert, M. G.M., Me.;Elsevier: New York, 1976; p 96. (37) Veith, J. A.; Sposito, G.Soil Sci. SOC.Am. J . 1977,41,870.

Langmuir, Vol. 7, No. 8, 1991 1705

Phosphate Adsorption on Boehmite 4. Conrtant-Capacitance Model. The constant-capacitance model is a surface-complexation model that assumes a ligand exchange mechanism for specific adsorption of ions at the oxidesolution interface.'* In the model allsurface complexes,including those formed with protons and hydroxyl ions, are inner-sphere complexes located in the same surface plane of adsorption; complex formation with ions from the background electrolyte is not included. The relationship between the surface charge, u(mol,L-9, and the surface potential, J. (VI, is assumed to be linear u = (CSa/F)J.

(1) where C (C V-I m") is the capacitance density parameter, S (ma g-1) is the specific surface area, a (g L-1) is the suspension density of the solid, and F (C mob-,-')is the Faraday constant. In the application of the model to phosphate adsorption on boehmite, the following surface reactions are defined:

+ H+(& = =AlOH,+,,)

(2)

=AlOH')(,) = =AlO-(,) + H+(.s)

(3)

=AlOH')(,)

=AlOH')(,)

+ H,PO,(,)

= =AIH(sr)POt-(,)

+ H,O + ZH+(.~) (4)

where x- is the charge on the phosphate surface complex ( x = 0, 1, or 2) and =MOHO(,) represents 1 mol of reactive surface hydroxyls bound to an Ala+ ion at the boehmite interface. The above reactions are described by the following intrinsic conditional equilibrium constants:

K+(ht)= exp(J.F/RT )[~AlOH,+]/([~AlOH')][H+I) (5) K-(ht, = exp(-+F/RT )([~AlO-][H+I)/[=AlOH"l

(6)

Ki(int)= exp(-xJ.F/RT )([=AlH(a-s)P04El [H+I")/ ([=AIOHol[H,PO,l) (7) where i = x + 1,R is the molar gas constant, and Tis the absolute temperature and brackets represent concentrations (mol L-1). Intrinsic conditional equilibrium constants are obtained by extrapolation of ordinary conditional equilibrium constants to zero net surface charge.'* Conditional equilibrium constants for adsorbed phosphate take the form

' K i = ([==AlH(2-s)P0,X-l[H+l')/([=AlOH')l [H,PO,])

(8) The mass balance equation for the surface functional group, =AlOH, is

+

[=AlOH], = [=AlOH')] [=AlOH:] + [=AlO-] + [=AlH,PO,"] + [=AlHPO;] + [=AlPO;-I

(9)

and the charge balance equation is

- [sAlO-] - (=AlHPO;] - 2 [=AlPO:-]

u 3 [==A10H2+]

(10) The above set of equations defines the equilibrium problem for phosphate adsorption on the aluminum oxide. The computer program FITEQLI was used to obtain the values of the intrinsic surface protonation-dissociation constants, log K*(i,,t), and the intrinsic surface complexation constants, log Kpi(m).m L c o n t a i n s the constant-capacitance model and used a nonlinear least-squares optimization technique to fit equilibrium constants to experimental adsorption data. Values of the intrinsic protonation-dissociation constants can be calculated from potentiometric titration curves of proton surface charge. In this analysis,lg it is assumed that the net surface charge, u, is equal to [mAlOH2+]at pH values below the point of zero net proton charge (PZNPC) and is equal to -[==AlO-] at pH values above PZNPC. The intrinsic conditional protonation-dissociation constants are related to the conditional protonation-dissociation constants ~

~~~

~

(38)Weatall, J. FITEQL: A program forthe determination of chemical equilibrium constants from experimental data, Re ort No. 82-01; Department of Chemistry, Oregon State University: 8orvallis, OR, 1982.

-3.0 2.0

4.0

8.0

6.0

10.0

12.0

PH

Figure 1. Proton surface charge of boehmite (yAl00H).

Aqueous suspensions contain lo00 ma L-l boehmite (187.2 m* gl-1) andO.OO1mol L-1 KCl. Open symbolsindicate experimental data (three experiments: circles, triangles, and squares), while the smooth curve is the fit by the constant capacitance model to the combined titration data.

as follows:

Kh(ht)= 'KKtexp(fJ.F/RT 1

(11)

After converting surface charge, u, to surface potential, J., using eq 1 and transforming both sides to common logarithms, the equation can be solved for log CKA log 'KKt= log

f up/ln(lO) CSaRT

(12)

By plotting log OKa versus u, log K*(ht)is obtained from the y intercept and the capacitance density parameter, C, can be extracted from the slope. The PZNPC, defiied in the constant capacitance model as PZNPC

(log K+(ht)- log K4ht))/2

(13)

can subsequently be obtained.

Experimental Results 1. Surface-ChargeProperties. The PZNPC for the pristine boehmite is about 7.2 (Figure l), lower than that reported for other aluminum The point of zero charge (PZC) for the pristine boehmite is about 9.4 (Figure 21, in close agreement with PZCs for boehmite and other aluminum oxides reported elsewhere.w*'O*Proton surface charge was measured by using the conventional titration technique; no provision was made to bring the boehmite suspension into solubility equilibrium during the course of the titration experiment. The suspensions used in the microelectrophoresisexperiments, on the other hand, were brought into solubility equilibrium through the addition of AlC13 prior to the mobility measurements. Adsorption of phosphate dramatically changes the zeta potential of the boehmite (Figure 2), lowering it from 9.4 to about 5.7. The effect of phosphate adsorption, as with (39) Yopps, J. A.; Fuerstenau, D. W. J. Colloid Sci. 1964, IS, 61. (40) Sadek, H.;Helmy, A. K.; Sabet, V. M.; Tadroe, T. F. J. Electroanol. Chem. 1970,27, 257. (41) Wie+ G.R.;Healy, T. W. J. Colloid Interface Sci. 1976,351,427. (42) Matuewc, E.Pure Appl. Chem. 1978,50, 1193. (43) Parks,G.A. Chem. Rev. 1966,6S, 177. (44)Alwitt, R. S. J. Colloid Interface Sci. 1972,40, 196. (45) Brace, R.;Matijevic, E. J . Inorg. Nucl. Chem. 1974,96,3691. (46) Anderson, M.A.; Malotky, D. T. J. Colloid Interface Sei. 1979, 72,413.

Bleam et al.

1706 Langmuir,Vol. 7, No. 8, 1991

(equivalent to 2.87 pmol m-2 or -57.9 A2 per site), the largest value of [=AlOH2+] occurring at pH = 3. Each titration curve was analyzed separately and the resulting log &(int) and C values were averaged log K+(int)= 5.6 f 0.2,logK-(int)=-8.6*0.2,andC= 1.8k0.9. ThePZNPC obtained by using eq 13 is equal to 7.1 and compares well with the crossover point at pH = 7.2 in Figure 1. Values for the intrinsic conditional protonation-dissociation constants were also obtained by fitting the potentiometric titration data with the FITEQL program. Because of the large standard deviation in the value of the experimental m c a capacitance density, C was set equal to 1.06 C V-l m-2, a N value used in a previous application of the constantcapacitance model to phosphate adsorption on aluminum , oxide minerals.& All titration data were analyzed by -40.0 -1 3 5 7 9 11 simultaneously. The resulting constants are log K+rmt)= 6.02 and log K+,t) = -8.45. The PZNPC obtained by F ~ PH of 7.24 is virtually identical with the crossover point of 7.2 and agrees well with the linear extrapolation value of 7.1. Figure 2. Zeta potential of pristine and phosphate-containing The fit of the constant-capacitance model to the potentioboehmite (r-Al00H) suspensions. Aqueous suspensionscontain metric titration data is excellent and is shown in Figure 260 m*L-'boehmite (187.2 m2g-') and 0.001 mol L-I KCl. Open squares indicate pristine suspensions; filled squares indicate 1. phosphate-containing suspensions. The fiial phosphate con2. Phosphate Adsorption Envelope. Figure 3 shows centrationin all phosphate-containingsuspensionswas 0.1 mmol the activities of the major solution-phosphate species as L-I, added as potassium dihydrogen phosphate. predicted by using the activity constants mentioned above. -2 The system consists of boehmite suspensions containing 250 m2 L-l with an ionic strength of 0.001 and a total phosphate concentration of 0.1 mmol L-l. The total Total AI concentration of aluminumis not, of course, constant since it follows the solubility of the boehmite. Aluminumphosphatesolution complexesare the dominant phosphate solution species at and below pH = 4. The actual adsorption envelope, measured by using the concentration-null method, is shown in Figure 4. Two characteristics of this adsorption envelope warrant comment. First, and most obvious, the adsorption envelope is sharply peaked at pH = 4, declining rapidly in the pH range 4 to 3. This form of adsorption envelope has been observed by both Chen et al. and Huang,b5*66 the adsorption maxima occurring at about the same pH in their studies as found in ours. Muljadi et al." report adsorption data 2 4 6 0 10 12 for both gibbsite and pseudoboehmite,which do not show an adsorption maximum at pH = 4. This phosphate PH adsorption behavioPgM is in contrast to that commonly Figure 3. Solution phosphate speciation in aqueous boehmite observed for goethite, for which the adsorption capacity (r-Al00H) suspensions. Total phosphate concentration equals continues to increase below pH = 4.1434*49157-59 0.1 "01 L-1. The second characteristic is that the pH of maximum other second-period o x y a n i ~ n s , is~to~reduce J ~ ~ the ~ ~ ~ ~phosphate adsorption we found is somewhat lower than magnitude of the positive charge on the surface. This that reported for other aluminum oxides.M*Qa Notice effect is well-known and has led some to suggest that this that the overall shape in the adsorption envelope (Figure occurs through the direct displacement by phosphate 4) is similar to the trend in the concentration of AlHPOd+(,) ligands of aquo and hydroxo ligands coordinating surface (Figure 3). metal cations.4140-62 It is interesting to compare our results with those of Values for the intrinsic conditional protonationAnderson and Malotky.4 They reported a lowering of dissociation constants and the capacitance density pathe PZC from 9.3 (pristine yAlzO3) to 5.6 (phosphaterameter were obtained by plotting the potentiometric "saturated" 7-AlrOs). Their PZC = 5.6 was determined titration data in the form of eq 5. The total number of at rmaxPhaPhate and compares with our measured PZC = reactive sites, [-A~OH]T, was set equal to 2.87 mmol L-' 5.7, also at I',(uphc*PbM (Figure 2). Their estimate of I',luPhmphate, measured by using adsorption isotherms, was (47)H b t o n , F.J.; Atkinaon, R. J.;Posner, A. M.;Quirk, J. P. Nature 2.1 pmol m-2 compared to our estimate, with 95% 1967,215,1459. confidence interval, of 1.8 f 0.3 rmol m-2 (Figure 4). (48) Bowden, J. W.;Nngarajah, S.; Barrow, N. J.;Posner A. M.;Quirk, J. P.A w t . J. Soil Re8. 1980,18,49. 80 01

I

I

.

1

\

I

(49)Barrow, N. J.; Bowden, J. W.J. Colloid Interface Sci. 1987,119,

233.

(50) Wan, S.S.S.;Perrott, K.

W.;Saunders, W.M.H.J . Soil Sci.

1974,25,438. (51)Wan, S. S.5.Nature 1976,262,45. (52)McLaughlin, J. R.;Ryden, J. C.; Byem, J. K.J. Soil Sci. 1981,32,

!le5 ---.

(63)Chon, Y.-S.;Butler, J. N.;Stumm, W.J. Colloid Interface Sci. 1973,43,421.

(54)Goldberg, S.;Sposito, G. Soil Sci. SOC.Am. J. 1984,48,772. (55)Huang, C. P. J. Colloid Interface Sci. 1976,53,178. (56) Muljadi, D.;Poener, A. M.; Quirk, J. P. J. Soil Sci. 1966,17,212. (57)Hinpton, F.J.; Atkinaon, R.J.; Posner, A. M.; Quirk, J. P. 2Von.s. 9th Int. Cong. Soil Sei. 1968,I, 669. (58)Hingeton, F.J.; Poener, A. M.; Quirk, J. P. J. Soil Sci. 1972,23, 177. - .

(59)Barrow, N. J.; Bowden, J. W.;Posner, A. M.; Quirk, J. P.A u t . J . Sod Res. 1980,18,395.

L

Langmuir, Vol. 7, No. 8, 1991 1707

Phosphate Adsorption on Boehmite

Table 11. Stoichiometry of the Equilibrium Problem: Phosphate Adsorbed at the Interface between Aqueous Solution and Boehmite (yA100H)

-AlOH(s) 3

species H+(aq) OH-(aq) =AlOHO(s)

&

c

E .

Was)

Als+(aq) H~P04~(aq)

HzPO4-(aq)

HP0d2-(aq) POP(aq)

KH2PO&aq)

V."

8

2

4

0

6

10

12

PH

a

KHPOl-(aq) KPOP(aq) AlOH2+(aq) Al(OH)z+(aq) Al(OH)ao(aq) Al(OH)4-(aq)

A12(OHh4+(aq) AlHzP042+(aq) AlHPOl+(aq) =AlOHz+(s)

AlO=AlH2PO4O(s) =mod*( S)

2

4

6

8

10

12

PH b Figure 4. Phosphate surface excess at the interface between

aqueous solution and boehmite (yAl00H) as a function of pH. Aqueous suspensions contain 0.1 m o l L-' phosphate, 250 m2 L-1 boehmite (187.2 m2 gl),and 0.001 mol L-l KCl. Phosphate surface excess was measured by the concentration-nullmethod. The smooth curves are the fit by the constant capacitancemodel to the experimental data: (a) without aluminum phosphate solution complexes; (b) with aluminum phosphate solution complexes. Clearly, the results we measured agree well with those observed by others. The constant-capacitance model was used to describe the experimental phosphate-adsorption data on boehmite. Previous applicationsof the model to phosphate adsorption on aluminum oxidesu have not included aluminumphosphate solutioncomplexes. The ability of the constantcapacitance model to describe phosphate adsorption was tested with and without the consideration of these species. The stoichiometry of the equilibrium problem defined for the model application including aluminum phosphate solution complexes is provided in Table 11. The total number of reactive sites, [=AOH]T, was set equal to 0.478 mmol L-l (equivalent to 1.91 pmol m-2 or =86.9 A2 per site), the maximum phosphate surface excess at pH = 4. The intrinsic conditional protonation-dissociation constants were set to the values obtained from linear extrapolation of the potentiometric titration data and C was set equal to 1.06 C V-1 m-2. In the initial analysis, the three phosphate surface complexation constants, log Kp$ht),were optimized by the FTIEQL program. This preliminary calculation indicated that log KP*(int), the phosphate surface constant for the

0 0 1 0 0 0

Al3+

0 0 0

0 0 0

0 0

0 0 0

1 0

0 0

0 0

0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0

1 1 1 1

H3-

K+

de+F/RT

0

0 0 0 0 1 1 1

0 0 0 0 0 0 0

0 0 0 0 0 1 -1

0 0

0

0

-2

0

0 0

0 1

0 0 0 0 0 0 0 1 1 1 1 2 1 1 0 0 0 0

PO4 H+

log K(bt)

0 1 0 . 0 0 0 -1 -14.00 0 0 0 . 0 0 0 0 . 0 0 0 0 0 . 0 0 1 0 0 . 0 0 1 -1 -2.15 1 -2 -9.35 1 -3 -21.70 1 -1 -2.20 1 -2 -8.20 1 -3 -19.80 0 -1 -6.02 0 -2 -9.30 0 -3 -14.99 0 -4 -23.33 0 -2 -7.69 1 -1 1.00 1 -2 -1.90 0 1 5 . 8 0 0 -1 -8.63 1 0 7.28 1 -2 -1.05

formation of =AlHPO,-(,), was unnecessary. Subsequently only two phosphate surface constants were optimized by the FITEQL program. The ability of the constantcapacitance model to describe phosphate adsorption on boehmite is indicated in Figure 4. The model underestimates adsorption at low pH and overestimates it at high pH. Interestingly, without consideration of aluminum phosphate solution complexes, the model results indicate no peak in the surface excess at pH = 4 (Figure 4a). Upon inclusion of these species, the model results describe an adsorption envelopethat, like the experimentaldata, peaks sharply at pH = 4 (Figure 4b). These results suggest that the observed decrease in phosphate adsorption at low pH need not be due to the minimum solubility of aluminum oxide as postulated by Chen et aleMbut simply occurs because solution phosphate concentrations are reduced by forming an aluminum-phosphate complex that does not adsorb. 3. 31PNuclear Magnetic Resonance Spectroscopy of Surface-Adsorbed Phosphate. The 31PNMR solution spectra of the supernatant solution from a suspension at pH = 3 appears in Figure 5. The resonance signal for uncomplexed solution phosphate should appear at =+0.8 ppm, yet the only two resonances appear at -8 and -10 ppm (the resonance at +30.73 ppm is a chemicalshift reference). These two resonancesundoubtedly arise from aluminum phosphate solution complexes, though their assignment is uncertain.32*"25 More importantly, the phosphate speciation appearing in Figure 3 appears to be qualitatively correct, viz. at low pH values there are negligible amounts of uncomplexes phosphate in our suspensions. Our solid-state, magic-angle-spinning (MAS)3lP NMR spectra of adsorbed phosphate were collected at five pH values: 4.0, 5.5, 7.0,9.0, and 11.0. These values include the pH of maximum adsorption (cf. Figure 4) and values corresponding to regions where the constant-capacitance model predictsMeither a single surface species dominates and (4.0 and 11)or two or more species coexist (5.5,7.0, 9.0), respectively. Single-pulse excitation MAS spectra at the pH values mentioned above are shown in Figure 6. There is a single,

1708 Langmuir, Vol. 7,No. 8, 1991

Bleam et al.

broad, upfield signal) covers a range of -12.7 ppm. The shielding of the & resonance decreases dramatically as the pH increasesand it is clearly more sensitive to changes in solution pH. Finally, the broadness and overall appearance of the & resonance strongly suggest that the phosphate species that give rise to this signal reside in a 11 distribution of chemical environments. The range of chemical environments we would associate with the Bo resonance is much narrower. Earlier we mentioned a special type of CP-MAS +@W. experiment, known as a “dipole suppression” or “inter1 1 1 1 I I I I I I I I I I I I I I I I I I I I I1 1 1 I ( 1 1 1 1 I J rupted-decoupling” e ~ p e r i m e n t . 2 This ~ ~ ~ experiment 40 30 20 10 0 -10 -20 -30 probes 1H-31P nuclear-dipole coupling. The results are Figure 5. 161.7-MHz*lPsingle-pulse-excitation,proton-decouuseful when studying hydrolysis, since it may be possible ’ phosphoric acid, scale pled solution NMR spectrarelative to 85 % to use them to discriminate whether or not a phosphate in ppm. The resonance at 30.73 ppm is from hexamethylphosis protonated. In a typical CP-MAS experiment the phoramide which was used as an externalchemicalshift reference. acquisition of the nuclear magnetic free-induction decay Supernatant solution ia from a phosphate-containing (total (FID) is initiated =l ps (viz. a 1 ps ”delay”)followingspinphosphate concentration equals 0.1 mmol L-9 aqueous boehlock and proceeds under conditions of high-power lH”’P mite (yA100H) suspension at pH = 3.0 12 ps for a 40’ pulse width; 10 OOO scans; 2k data points zero-filled to 16k; 16-kHz decoupling. In an “interrupted-decoupling” experiment, spectral width; 0.16-8 recycle time. the start of FID acquisition, as well as 1H-31P decoupling, is delayed for some interval of time following spin-lock very broad resonance in the pH = 4 spectrum and two during which the 31Psignal intensity is permitted to decay signals in the spectra at all other pH values: a narrow, under the influence of 1H-31P coupling. The extent of downfield resonance (Bo)and a broad, upfield signal (&I. signal loss becomes a measure of 1H-31P nuclear dipole There are clearly several unresolved resonances in the A0 coupling: the more rapid the signal loss the greater the peak. The spectra at pH = 5.5 and pH = 7.0 are quite coupling. similar to one another, as are the spectra at pH = 9.0 and In a previous study,m we found that in those phosphate pH = 11.0 similar to each other. minerals where crystal structure refinements indicate no It appears that at a third “peak” is present in Figure protonation of the phosphate molecule (e.g.,variscite, cran6b,c at =20 ppm and in Figure 6e a shoulder at -5 ppm. dallite, wavellite)the 3lP resonanceintensity is still -20% The signal-to-noiseratio of the =20 ppm “peak” (Figures of its initial value after interrupting lH”lP decoupling 6b,c) is very low and we cannot be certain that it truly 250 ps for resonances. The 31P signal intensity is comrepresents a distinct chemical environment. Although the pletely lost after the same interrupt interval in those signal-to-noiseratio of the -5 ppm shoulder in Figure 6e minerals where protonation of the phosphate does occur is higher than the -20 ppm “peak”,none of these represents (e.g., KDP). a significant proportion of the phosphate surface excess. 31PCP-MASspectra of surface-adsorbedphosphate are More importantly, there is simplytoo little signal intensity shown in Figures 8-11, two spectra at each pH value. One to make any interpretations about the chemical environis a typical CP-MAS spectrum, =l ps delay between the ments represented by such weak signals. end of spin-lock and the beginning of FID acquisition. The pH dependence of the isotropic chemical shifts of Henceforth, we will refer to these simply as CP-MAS the major NMR signals appearing in the spectra of Figure spectra. The other is an ‘interrupted-decoupling” CP6 are illustrated in Figure 7. Only the major signals are MAS spectrum with a 250-psdelay. We will refer to these included in this graph for the reasons just given. as “dephased” CP-MAS spectra. Without exception the In preparation of Figure 7, it seemed reasonable to & signal is almost completely decayed in the udephased” connect the downfield Bo resonances of the 5.5 and 7.0 CP-MASspectra while significantsignal intensity remains spectra to one another and to draw a line connectingthose from the Bo signal. Considering the signal-to-noiselevel in the 9.0 and 11.0 spectra. If we do the same for the in the “dephased” CP-MAS spectrum at pH = 4.0 (Figure upfield & resonances, connecting 5.5 to 7.0 and 9.0 to 8),it is difficult to assign signal intensity accurately, yet 11.0,a pattern appears which suggests that the chemical it is less than 10% of the CP-MAS intensity. For thls shift of the Bo resonance varies continuously from =O.l reason, we interpret this result as evidence that the & ppm at pH = 5.5 to -3.9 ppm at pH = 11.0 and the A0 resonance arises from a phosphate with protonated oxresonances vary continuously from --11.2 ppm to -1.5 ygens. ppm in the same pH range. The pH-dependent 31P There is another point to be made concerning these chemical shift of solution phosphate is also plotted in spectra. First-order spinning side bands (ssb) are quite Figure 7. noticeable in the CP-MAS spectra at every pH for both Although the Bo resonance varies over a chemical-shift up- and downfield signals. The ssb of the downfield Bo range that is about the same as that covered by solution signal is more noticeable primarily because the resonance phosphate (0.8ppm at pH = 4 to 3.4 ppm at pH = 111, is much narrower than the upfield &signal. The intensity clearly the Bo resonance of adsorbed phosphate and of the ssb in a material free of paramagnetic ions is related solution phosphate display different pH-dependent chemto the chemical shift anisotropy of the nuclei associated ical-shiftbehavior. The A0 resonancecovers a much larger with that resonance.@-’ A rapidly tumbling phosphate chemical-shift range, displaying a great sensitivity to molecule would produce a narrow resonance, but it would solution pH. not have significant ssb intensity since rapid tumbling In summary, the pH-dependent chemical shift data of would tend to average to zero the anisotropic contribution these spectra suggest that the surface-adsorbedphosphate to the chemical shift. can be grouped into two classes. The chemical shift of one (givingrise to the narrow, downfield Bo signal) covers (60)Bleam, W. F.;Pfeffer, P. E.; Frye,J. S.Phys. Chem. Miner. 1989, a range of -3.8 ppm, while the other (the source of the 16, 455.

1

dj*.

Langmuir, Vol. 7, No. 8, 1991 1709

Phosphate Adsorption on Boehmite

d

Fi ure 6. 121.49-MHz SIP single-pulse-excitation solid-state NMR spectra (85% HaPo4 external reference, scale in ppm): ==3kHz S spinning frequency, 256 scans, 35.7 kHz spectral width, 5 s recycle time. Freeze-dried boehmite with surface-adsorbed phosphate: (a) pH = 40,128 data points zero-filled to 8k; (b) pH = 5.5,128 data points zero-filled to 8k; (c) pH = 7.0,128 data ponta zero-filled to 8k; (d) pH = 9.0,256 data points zero-filled to 8k; (e) pH = 11.0, l k data points zero-filled to 8k. Main peaks are labeled with the subscript "0", first-order spinning sidebands with the subscript 'fl".

Id

in

o

".

2

4

8

6

10

12

PH Figure 7. chemical shift as a function of pH. Squaresindicate the single-pulse-excitation chemical shift data for phosphate adsorbed on the surface of boehmite (cf. Figure 6). Filled squares correspond to the peaks labeled 'Ao" while open squares indicate the 'Bo" resonance. Circles indicate the chemical shift NMR data for 0.1 mmol L-*potassium phosphate solutions. We estimated the relative intensities of the up- and downfield signals by determining t h e relative peak areas from single-pulse MAS spectra (Figure 6). We find that the surface species giving rise t o t h e upfield A0 signal determines the p H dependence of the phosphate surface excess, while the surface excess of the phosphate associated with the downfield Bo resonance is more or less constant in t h e p H range from 5.5 t o 11.0 (l'Phapbb*ddield = 0.27

s Figure 8. 121.49-MHz cross-polariition solid-state NMR spectra (85% ' H 8 0 4 external reference, scale in ppm): 2S6acana; 128 data points zero-filled to 8k; 35.7-kHz spectral width, 6-8 recycle time; 600-rs contact time. Freeze-dried boehmite with surface-adsorbed phosphate from a pH = 4.0 suspension, delay between spin-lock and acquisition: 1ws (upper); 250 pa (lower). Main peaks are labeled with the subscript "0",fit-order spinning sidebands with the subscript 'fl". f 0.07 pmol m"). We could not include p H = 4 in this analysis because the signal is too broad to resolve t h e Bo resonance. T h e surface excess of the phosphate responsible for Bo resonance would be equivalent to adsorbing 4 7 f 4% of t h e total solution phosphate. Another way to visualize this result is to imagine t h a t if t h e interstitial solution (which contains 0.1 mmol L-1 KDP) in t h e solids after

Bleam et al.

1710 Langmuir, Vol. 7, No. 8,1991 Bo A,

c

0

Figure 9. 121.49-MHz S I P cross-polarization solid-state NMR spectra (85 % HPO, external reference, scale in ppm): 256 scans, 256 data points zero-filled to 8k, 35.7-kHz spectral width, 5-8 recycle time, 600-ps contact time. Freeze-dried boehmite with surface-adsorbed phosphate from a pH = 5.5 suspension, delay between spin-lock and acquisition: 1 @ (upper); 250 (lower). Main peaks are labeled with the subscript ‘On, first-ordering spinning side-bands with the subscript ‘fl”.

2

a

6

4

a

10

12

PH 100

BO

a,

a

01 2

Figure 10. 121.49-MHz nP cross-polarization solid-state NMR spectra (85 % H@Od external reference, scale in ppm): 256 scans, 256 data points zero-filled to 8k, 35.7-kHz spectral width, 5-8 recycle time, 600-ps contact time. Freeze-dried boehmite with surface-adsorbed phosphate from a pH = 9.0 suspension, delay between spin-lock and acquisition: 1 ps (upper); 250 ps (lower). Main peaks are labeled with the subscript “0”,first-order spinning side-bands with the subscript ‘fl”. BO

-

Figure 11. 121.49-MHz SIP cross-polarization solid-state NMR

spectra (85% Hap01 external reference, scale in ppm): 256 scans, 400 data pointa zero-filled to 8k, 35.7-Wz spe&ral width, 6-8 recycle time, 600-ps contact time. Freeze-dried boehmite with surface-adsorbed phosphate from a pH = 11.0 suspension, delay between spin-lock and acquisition: 1 @ (upper); 250 ps (lower). Main peaks are labeled with the subscript “O”, fit-order spinning side-bands with the subscript ’fl”. centrifuging constituted 17 i 4% of the total suspension

volume ( 4 . 5 mL in our experiments since we used 50-mL suspension volumes), then freeze-drying the centrifugate would force the precipitation of this amount of interstitialsolution phosphate.

b

,

! 4

.

.-1 0

6

i

I

10

12

PH

Figure 12. Surface speciation predicted from fitting the constant capacitance model to the titration and phosphate adsorption data appearing in Figures 1 and 4: (a) without aluminum phosphate solution complexes; (b) with aluminum phosphate solution complexes.

Discussion 1. Hydrolysis of Phosphate Surface Species. The

phosphate surface speciation predicted by the constantcapacitance model using the average phosphate surface complexation constants for varioue aluminum oxides (log &‘(int) = 9.34,log Kp2(ht)= 3.34,log = -2.87)given by Goldberg and SpositoM is indicated in Figure 12a. The species =AlH2PO4O(,) dominates below pH = 7.5 and the species =AlP042-(,) dominates above pH = 8. The species =dHPO4-(,) is of secondary significance since it never constitutes more than 40% of the phosphatesurface excess. Phosphate speciation on the surfaceof boehmitepredicted by the constant-capacitance model in the presence of aluminum-phosphate solution complexes is shown in Figure 12b. The model predicts that =AlH2PO& is the dominant phosphate surface species only up to pH = 4.5. A t higher pH values, the nonprotonated species =A1P042-(8) dominates. We designed a series of 3lP solid-stateNMR experiments to test in part the surface-speciationpredictions made by the constant-capacitance model. We recognize that ‘dthough surface-complexation models give the appearance of having a definite microscopic basis, Spositoel derived (61) S p i t o , G.J. Colloid Interface Sci. 1982, 91, 329.

Phosphate Adsorption on Boehmite

Langmuir, Vol. 7, No. 8, 1991 1711

a phenomenological model with the same mathematical structure. Still, this does not mean that speciation predictions are not valid. Before discussing the NMR results, it would be helpful totranslate the picture drawn by the constant-capacitance model into the language of NMR. Whether the surface speciation is as illustrated in Figure 12a (viz., some combination of all three species: rA1H2P04°(s), =AlHP04-(,), and =AlP042-(,)) or as in Figure 12b (viz., a combination of only two species: =AlH2PO4O(,) and =AlP042-(,)), we would expect only one NMR resonance in the spectrum arising from these surface species because they are likely to be undergoing rapid chemical exchange. Our reasoning is the following. 31PNMR spectra of aqueous HsPOd-MOH solutions (M = Li, Na, K, Cs) show only one resonance because of chemical exchange involving the various phosphate species. If the lifetime of two or more species is short on the time scale of an NMR experiment, then a single peak will appear at a chemical shift that is the weighted average of the chemical shifts of those species undergoing chemical exchange. The pH dependence of this time-averaged resonance simply reflects changes in the proportions of the various species as they hydrolyze. In contrast, the two signals at -8 and -10 ppm in 3lP solution spectra of Figure 5 represent chemical enwronments that are not in chemical exchange with one another on the time scale of the NMR experiment. Surfacespeciessuch as =5AlH2P04°(,), =AlHPo4-(8),and =AlP042-(,) differ only in the number of protons bound to the phosphate. Protons exchange rapidly in solution and, we would also expect, at an oxide/aqueous solution interface with the result that surface species of this type would be in rapid chemical exchange and give rise to only one resonance. Although our samples were freeze-dried, no effort was made to remove all water from them, and surfacehydration would be sufficientto provide a pathway for proton exchange. The appearance of more than one peak in the spectrum means that more than one chemical environment exists and these are not in rapid chemical exchange with one another. Arguing by analogy with the picture of phosphate chemical exchange in solution, these distinct chemical environments probably differ by more than simply the number of protons bound to the phosphate. The speciation given in Figure 12a predicts that protonated phosphates would be significant as high as pH 9, while the speciation illustrated in Figure 12b predicts that protonated phosphates would be significant only as high as pH 6. "Dephased" CP-MAS results should indicate that pH at which this critical transition occurs. Dipolar coupling between the 31Pnuclei of deprotonated surface phosphates and 'H on the oxygens coordinating the aluminum atoms should still be sufficient to cross-polarize. In short, we should still expect to see a resonance in the CP-MASspectra even when EA~PO~~-(,) is the only surface species unless desorption of protons from the boehmiteat high pH effectivelystrips the surface clean of protons. Our results indicate that the pH dependence of the adsorption envelope is due primarily to a population of phosphates which we associated with the & signal in our NMR spectra. The breadth and overall appearance of the & signal suggests that it is generated by phosphate molecules existing in a range of chemical environments. A 7.0-T field strength is not sufficient to resolve distinct resonances in the & signal. This lack of resolution does not prevent us from drawing a number of valuable conclusions concerning the chemical

-

-

environment of these phosphates. First, the chemical environment is directly influenced by the pH of the suspension. This comes as no surprise since we know that the surface charge and the affinity of the boehmite surface for adsorbing protons and hydroxyl ions is continuously changing with pH. Second, we interpret the "dephased" CP-MAS spectra (Figures 8-11) as proof the oxygens of those phosphates responsible for the & signal are directly protonated. From this we conclude, there are still significant amounts of either =AlH2PO4O(,) or =AlHPO4-(,) at pH = 9 and the transition to the point where =AlP042-(,) dominates must occur between pH = 9 and pH = 11. Finally, desorption of protons from the surface of boehmite is sufficient to render lH-9lP cross-polarization impossible at pH = 11. 1H-31P cross-polarization does occur in wavellite where there are water molecules and hydroxyls coordinating aluminum atoms but none of the phosphate oxygens appear to receive hydrogen bonds.@) Exactly how extensive deprotonation of the oxide surface must be in order to eliminate cross-polarizationis not clear. 2. Surface Complexes and Surface Precipitates. An adsorbate may be in one of three major chemical environmentsat the mineral/ aqueous solution interface: outer-sphere surface complexes, inner-sphere surface complexes, and surface precipitates. NMR experiments can provide very useful information about these chemical environmenta and their response to solution-pH changes. Hydrolysis in solution or at a mineral/solution interface represents a change in chemical environment observed as a pH-dependent chemical shift. Both inner- and outersphere surface complexes should exhibit pH-dependent chemical shifts (vide supra),distinguishingthese chemical environments from surface precipitates. The major property distinguishing inner- from outersphere surface complexes is their dynamics. Outer-sphere surface complexes are expected to display "solution-like" mobility, thereby averaging the principal components of the chemical-shifttensor. Spinning side bands would not be associated with solid-state NMR resonances of outersphere surface complexes. A rapidly tumbling outersphere complex would also be difficult to cross-polarize and may not appear in a CP ~pectrum.6~+3~ An innersphere surface complex, on the hand, would most likely exhibit significant ssb intensity and ready cross-polarize. Thus, because the upfield & signal displays ssb intensity and readily cross-polarizes, we conclude & represents inner-spherephosphate complexes with surfacealuminum atoms. Surface crystallites or precipitation nuclei could represent a chemical environment with the properties we observe for the narrow Bo resonance. With the exception of certain crystalline materials synthesized by dissolving aluminum metal in phosphoric acid and the mineral taranakiteF2there are no aluminum phosphates that we are aware of containing protonated phosphates in the crystalstructure.M The Bo resonance behaves as if it were not protonated yet in sufficientlyclose proximity to protons (possibly those in water molecules or hydroxyl ions coordinating the same aluminum to which the phosphate itself is bonded) to yield a resonance in a CP-MAS spectrum. The narrow peak width suggests a much more restricted chemical environment, while the intensity of the ssb supports the notion that the phosphates in this population are not motion averaged. (62) Bleam, W. F.;Pfeffer, P. E.; Frye, J. 5.phy8. Chem. Miner. 1989, 16,809.

(63) Bank, S.; Bank, J. F.; Ellie, P.D. J. Phye. Chem. 1989,93,4847. (64) Kniep, R. Angew. Chem., Znt. Ed. Engl. 1986,24525.

1712 Langmuir, Vol. 7, No. 8,1991

Such a crystallite would have to be small enough so that allphosphate anionsin the nuclei would respond to changes

in surfacecharge. The chemicalenvironmentat the surface is continuously changing with pH, as evidenced by the proton surface charge (Figure l),the zeta potential of the suspension particles (Figure 2), and the chemical shift (Figure 8). Weiss et a1.& observed a dependence of isotropic chemical shift on surface charge in the 29SiNMR spectra of phyllosilicates. Though the surface charge in phyllosilicatesarises from isomorphoussubstitution rather than protonation, it may be possible that the pH dependence of the 31Pchemical shift we are observing in the Bo resonance arises from a variation of surface charge. The phosphate surface excess on aluminum oxides at pH = 4 generally falls in the range of 2-3 pmol m-2.M@*ss It would be fortuitous that these particular values would appear in at least three separate studies besides our own, unless a well-defined adsorption phenomenon were occurring. Phosphate removal from solution plateaued following repeated additions of phosphate in all of our "concentration null" experiments. If precipitation were occurring, we should not have seen a saturation of our adsorbing capacity through such an experiment. We conclude that we are, in fact, observing a well-defined adsorption phenomenon for which "surface-excess maxima" has meaning. Until NMR spectra are collected on moist samples, the status of the Bo resonance as a form of surface phosphate (65) Weiea, C. A.; Altaner, 5.P.;Kirkpatrick,R. J. Am. Mineral. 1987, 72,935.

Bleam et al. remains open. Our results are insufficient to conclusively rule out precipitation as an artifact of freeze-drying.

Conclusions The properties of the broad & signal confirm key elements of the hypothesis we were testing, viz. innersphere surface complexes form between phosphate and the aluminum atoms of the boehmite surface and the phosphates in these complexes hydrolyzewith varying pH. Our results will not permit direct confirmation of the specific species that form, their stoichiometry, or the explicit form of binding. The resolution of spectra from surface-adsorbed phosphate obtained by using the most advanced IR techniques, employing diffuse- and internal-reflectance cells and computer subtraction of background spectra, make detailed band assignments very difficult.Mt67 Yet, recently, Tejedor-Tejedor and Andersonala have presented evidence for the protonation of phosphate adsorbed on the surface of goethite. Acknowledgment. We authors are grateful to Tom Boswell, Eastern Regional Research Center. W.F.B. wishes to thank summer interns Amy Saunders and Cliffae Wallace, who assisted with the adsorption experiments. (66)Nanzyo, M.J. Soil Sci. 1984,35,63. (67) Nanzyo, M.; Watanabe, Y. Soil Sci. Plant Nutr. 1982,28, 369. (68) Tejedor-Tejedor,M. I.; Anderson, M. A. Langmuir 1986,2,203. (69) Tejedor-Tejedor,M. I.; Anderson, M. A. Langmuir 1990,6,602.